Obstacle Avoidance via Landmark Clustering in a Path-Planning Algorithm

In this paper we present a new 2D decentralized path-planning algorithm for a swarm of multi-rotor UAVs operating in an unknown environment. The way-points of the reference trajectories are computed as solutions of a sequence of optimization problems. In order to obtain coordination, the optimization problems are defined by considering the goal of the flight mission, the desired formation shape and the detected obstacles. The obstacle avoidance strategy is based on the clustering of the detected landmarks. Then the no-fly zones are obtained by fitting the minimum area rectangle boxes surrounding the clusters. The algorithm is tested through simulations of realistic scenarios.